OmniHallu: Unified Hallucination Detection for Cross-Modal Comprehension and Generation in Multimodal Large Language Models
OmniHallu is introduced, a unified hallucination detection framework spanning both comprehension and generation tasks across image, video, and audio modalities, and a preference-optimized trainable verifier that approximates the multi-agent decision boundary, reducing expert calls by 66% with minimal performance loss.
JianJiang Yang, Pei-Hang Li, Shan Xu et al.
· 0 citations